context-cleanup_skill

Overview

This skill analyzes and archives low-value memory notes in an OpenClaw workspace to reduce context bloat and improve agent responsiveness. It follows a safe, user-confirmed workflow that favors archiving over deletion. The tool is designed to be conservative with important files and to produce machine-readable outputs for automation.

How this skill works

The skill inspects workspace memory logs and applies a configurable policy to identify low-value or redundant entries. It generates a non-destructive plan listing archive candidates, asks for user confirmation, and then moves selected entries to an archive location rather than permanently deleting them. Outputs can be produced in human-readable or JSON formats for integration into CI or maintenance scripts.

When to use it

• When the agent workspace has grown large and response times are degrading.
• Before major model or workflow upgrades to reduce noisy context.
• As part of routine maintenance to keep memory focused and relevant.
• When preparing an environment for backup or snapshot to minimize size.

Best practices

• Always run the analyze and plan steps before archiving to review candidates.
• Use the dry-run option to preview changes without modifying data.
• Prefer archiving over deletion to retain historical data for audits.
• Exclude core files and folders (e.g., persistent specs and main memory files) from automated cleanup.
• Use machine-readable outputs (JSON) when integrating into automation pipelines.

Example use cases

• Analyze workspace memory to find redundant conversation fragments and archive them.
• Run a monthly maintenance job that plans cleanup and requires human approval before archiving.
• Dry-run an archive step to estimate reclaimed context size ahead of a major deployment.
• Automate cleanup in a staging environment using JSON outputs to feed a CI task.